This PDF is a selection from a published volume from the
National Bureau of Economic Research
Volume Title: Analyses in the Economics of Aging
Volume Author/Editor: David A. Wise, editor
Volume Publisher: University of Chicago Press
Volume ISBN: 0-226-90286-2
Volume URL: http://www.nber.org/books/wise05-1
Publication Date: August 2005
Title: Caste, Culture, and the Status and Well-Being of Widows
in India
Author: Robert T. Jensen
URL: http://www.nber.org/chapters/c10366
11
Caste, Culture, and the Status and
Well-Being of Widows in India
Robert Jensen
11.1 Introduction
Issues in aging are becoming of increasing importance in India. While
this may seem paradoxical because of the low life expectancy (about sixtyfour years for women and men), these numbers are somewhat misleading
because they reflect in part very high infant mortality rates. Conditional on
surviving to age sixty, male life expectancy in India is about sixteen years
for men and seventeen years for women (Irudaya Rajan, Mishra, and
Sankara Sarma 1999), which is not much less than the eighteen to twenty
years for most developed countries. India currently has the world’s second
largest population of elderly persons, with over 70 million persons aged
sixty or above (though India still has a relatively young population, in that
this represents only about 7 percent of the total population). This population is projected to grow to about 160 million within two decades, which
will represent about 12 percent of the total population at that time. Thus,
the elderly present an extremely large and rapidly growing population in
India.
Research on the elderly in India has increasingly focused on the wellbeing of widows in particular (Chen 2000; Drèze and Srinivasan 1997;
Chen and Drèze 1995). About 55 percent of women aged sixty and above
are widowed, compared to about 15 percent widowers among men (Irudaya Rajan, Mishra, and Sankara Sarma 1999); this difference reflects in
Robert Jensen is an associate professor of public policy at the Kennedy School of Government, Harvard University, and a faculty research fellow of the National Bureau of Economic
Research.
The support of the National Bureau of Economic Research (NBER) Center for Aging and
Health Research for this project is graciously acknowledged.
357
358
Robert Jensen
part the large husband-wife age gap (six years) as well as a greater incidence of remarriage among widowers compared to widows. Widowhood
is generally thought to represent a particular economic vulnerability for
women. In rural areas, most earnings opportunities involve strenuous
physical labor, such as in agriculture, which becomes increasingly difficult
with age. Caste norms also often discourage women, and widows in particular, from working, and in many cases women face difficulties in retaining control of their husband’s land upon his death. Further, private pensions are not common, and there is little in the way of public support for
the elderly.1 Drèze and Srinivasan (1997) find that, even with a range of
adjustments for equivalence scales, widows are no more likely to live in
poverty than nonwidows, but because widows tend to live in households
with fewer members, even small adjustments for economies of scale result
in widows’ being significantly worse off than average. However, almost all
studies of the well-being of the elderly are unable to examine the distribution of resources within households, using household income or expenditure per capita as a proxy for the resources available to the elderly person,
and there is ample evidence to suggest that widows often do not receive an
equal share within the household (Chen 2000).
Beyond the economic difficulties faced by widows, authors have also
pointed to social, cultural, and political limitations widows often encounter. For example, Chen (2000) points out that social mores often discourage widows from remarriage and dictate changes in their diet and behavior, and widows are often unwelcome at social events and religious
festivals and avoided by others because they are considered bad luck. More
generally, there are a variety of customs, norms, practices, beliefs, and institutions that affect the economic, social, and political opportunities and
the social status of the elderly, especially widows (the most infamous,
though perhaps least widespread, of which is sati, or self-immolation on
the husband’s cremation pyre). These practices are important not just because of the economic constraints they may place on widows, but in their
own right as well, as part of a broader conception of well-being that expands beyond purely economic terms (e.g., Sen 1999).
In this paper, we examine how these norms, attitudes, and practices vary
1. In principle, public support for the elderly is a joint responsibility of the state and national governments. There is a national pension scheme for impoverished elderly persons
aged sixty-five or older. Coverage is not extensive, eligibility is limited to only persons with no
surviving sons and below a (very low) income threshold, and states have statutory maximum outlays. As a result, very few elderly persons participate. For example, Chen (2000) finds
that only 10 percent of widows in her sample, and 23 percent of those with no adult sons, receive pensions. Pension levels are low, typically about 100 rupees (about U.S. $2) per month,
though some states have supplemental programs. There is a mandatory pension system for
firms with more than twenty employees and a separate system for civil servants. However,
about 90 percent of all workers are in the uncovered, informal sector (though a few states have
implemented limited schemes for this sector).
Caste, Culture, and the Status and Well-Being of Widows in India
359
across groups in India and how they affect the well-being of the elderly. We
first examine the influence of caste on well-being. Caste is a social organization prevalent in India that creates a well-defined social ordering (see appendix for a brief note about caste). Caste is highly correlated with other
measures of socioeconomic status (SES), with those from higher castes
typically better off along most dimensions than those from lower castes.
Castes also differ in terms of behavior, practices, and norms; most relevant
for the present case, higher castes often place more restrictions on and assign lower status to women, and widows in particular. Thus, caste offers
two potentially opposing influences on the well-being of women. Our objective is to examine the net consequences for widows, using indicators for
individual nutritional status and health rather than relying on household
income or expenditure per capita.
The second theme of this paper is examining whether social and cultural
institutions and the status of widows have any underpinnings in economic
factors. In particular, we explore whether the presence in the village of residence of crops for which women and the elderly are likely to have a comparative advantage, namely those that are less strength intensive (i.e., crops
not based on draft plough agriculture), means that women and the elderly
are able to make larger economic contributions (and have improved opportunities outside the household) and thus the status and treatment of
widows, and women more generally, will be more favorable. This exercise
also carries larger lessons about the potential role of economic factors in
the determination and evolution of social or cultural institutions and practices.
11.2 Widowhood and Remarriage
Before beginning the analysis, it is useful to explore the broad patterns
relating to widowhood and remarriage across various groups. To do so,
we make use of data from the 1998–99 National Family Health Survey
(NFHS),2 a nationally representative survey of 90,303 ever-married
women aged fifteen to forty-nine (which can be weighted to generate statistics representative at the state and national levels). The survey collects
data on a range of demographic, social, and (to a lesser extent) economic
variables for the household and its various members, as well as a villagelevel survey with information on agricultural, economic, and social conditions and infrastructure. These data do not contain a great deal of detail on
elderly persons, but data from the household roster can be used to describe
the patterns of widowhood.
2. The NFHS is the Indian implementation of the Demographic and Health Surveys (DHS;
now Measure DHS). The data, questionnaires, and additional information are available at
http://www.measuredhs.com.
360
Robert Jensen
Fig. 11.1
Current widowhood by age
Source: NFHS 1988–89.
Figure 11.1 shows NFHS data on the incidence of current widowhood
by age for men and women. At all ages the incidence is much higher for
women than men. About 40 percent of women aged sixty are widowed
compared to 12 percent of men, and the gap grows to 69 and 19 by age seventy. The large disparity at each age is the result of the large average husband-wife age gap (six years) and the greater incidence of remarriage
among men.3 Despite popular perception, remarriage is not prohibited by
any caste or religious group in India (Chen 2000). Unfortunately, the
NFHS only collects more information on marital histories for female respondents aged fifteen to forty-nine, including whether they have had a
previous marriage (but not separately indicating whether the first marriage
ended due to marriage or divorce).4 We focus on women aged forty-five to
forty-nine, to capture the oldest population available. Table 11.1 shows
data on the incidence of widowhood and remarriage, disaggregated by religious and caste groupings. The data reveal that 15 percent of women of
this age have ever been widowed or divorced, and 11 percent of everwidowed/divorced women are currently married. The greatest incidence of
ever-widowhood is among women of scheduled tribes, at nearly one in five.
3. Very few data sets, and no national-level data sets, collect information on remarriage.
However, a few studies for specific villages or regions have collected this information. For example, Chen’s (2000) sample reveals a remarriage rate of 10 percent among widows, while
Drèze (1990) finds 15 to 20 percent. Agarwala (1970) finds that 60 percent of widowers in
Delhi remarry, compared to 37 percent for women, and United Nations (1961) estimate the
rates at 60 and 10 percent respectively in rural Mysore (Karnataka). Bhat and Kanbargi
(1984) estimate 63 and 34 percent remarriage rates for men and women across all India, using information on current marital status and applying a cohort-based inferential technique
to census data.
4. However, divorce is not very common in India, significantly less so than widowhood.
Caste, Culture, and the Status and Well-Being of Widows in India
Table 11.1
361
Widowhood and remarriage, women aged 45 – 49 (%)
Ever widowed
Ever widowed and remarried
Currently widowed
All India
.15
(.01)
.11
(.01)
.13
(.01)
Scheduled tribe
.18
(.01)
.16
(.01)
.15
(.01)
.12
(.01)
.15
(.01)
.18
(.02)
.21
(.03)
.11
(.02)
.11
(.02)
.07
(.01)
.17
(.03)
.19
(.04)
.14
(.01)
.14
(.01)
.13
(.01)
.11
(.01)
.13
(.01)
.15
(.01)
Scheduled caste
Other backward caste
Forward caste
Muslim
Christian
Source: Author’s calculation from the NFHS.
Notes: Standard errors in parentheses. The NFHS data only indicate whether a woman has
been previously married, and do not indicate whether the first marriage ended due to widowhood or divorce. However, divorce is very uncommon in India.
Reflecting underlying economic differentials, there is a gradient by caste,
with the higher (forward) castes having only two-thirds the widow rate of
the scheduled tribes. However, the rate of remarriage is significantly higher
for the scheduled tribes, at nearly three times the rate among the higher
castes, and almost twice that of scheduled castes and other backward
castes (OBCs). Thus, in terms of current widowhood, there is not a significant gradient by caste. Muslims and Christians have similar widowhood and
remarriage rates as the scheduled tribes.
For Hindus, within castes,5 two general factors are thought to affect the
likelihood of remarriage. First, the likelihood of remarriage (and social approval of remarriage) declines with age, and becomes extremely rare after
age forty. An additional important factor is whether the woman has had
any children, in particular sons, in her first marriage. In general, even controlling for age, those who have no children and especially no sons will be
more likely to remarry. This distinction is motivated in part by choice (for
example, a woman may fear a new husband would treat her children
poorly, or she may feel less of a need to remarry because the son can help
support her) and in part by constraint (no one will remarry a widow with
children, or society may not condone it). The NFHS data reveal that conditional on having been ever widowed or divorced, having had a child from
5. There is a limited notion of caste among Muslims in India, even more limited among
Christians. However, caste for non-Hindus is often much less overt and does not connote the
same divisions as among Hindus.
362
Robert Jensen
the first husband decreases the likelihood of remarriage by about 43 percent, and conditional on having had a child, women with at least one son
are 23 percent less likely to remarry than women with only daughters (results not shown).
11.3 Socioeconomic Status, Caste, and Well-Being
Given the widespread incidence of widowhood, especially among older
women, it is important to examine the well-being of widows. Unfortunately, the NFHS provides only very limited data on SES; while there is
some information on employment status and assets owned, there is no information on income or expenditure. Further, even if household income
were available, the measurement of poverty among the elderly is very sensitive to assumptions on economies and scale and costs of children
(Deaton and Paxson 1998), as well as assumptions on the intrahousehold
allocation of resources. However, as an initial investigation, the NFHS
data do provide information on whether individuals work, either outside
the household for pay or in self-employment. Table 11.2 provides information on female employment for both widows and nonwidows aged fortyfive to forty-nine, with a distinction drawn between land-owning and landless households. Female employment rates among nonwidows are low
overall, at 20 percent among those owning land and 40 percent among the
landless. In general, employment rates are higher among the scheduled
tribes and castes relative to the OBCs and forward castes (even though land
ownership is highest among the scheduled tribes). For all castes and reliTable 11.2
Employment among women aged 45–49, by widowhood status, NFHS
Nonwidows
Widows
Work
for pay
Own
land
Work for pay
(landless)
Work
for pay
Own
land
Work for pay
(landless)
All India
.20
(.01)
.52
(.01)
.40
(.02)
.39
(.01)
.43
(.01)
.48
(.05)
Scheduled tribe
.29
(.01)
.31
(.01)
.21
(.01)
.14
(.01)
.11
(.01)
.33
(.01)
.72
(.01)
.44
(.03)
.58
(.01)
.52
(.01)
.42
(.01)
.47
(.01)
.43
(.02)
.35
(.01)
.26
(.01)
.24
(.01)
.14
(.01)
.41
(.02)
.53
(.04)
.50
(.03)
.43
(.02)
.30
(.02)
.23
(.03)
.43
(.04)
.60
(.04)
.34
(.03)
.45
(.02)
.46
(.02)
.31
(.03)
.36
(.04)
.54
(.05)
.43
(.01)
.47
(.03)
.32
(.02)
.30
(.04)
.48
(.06)
Scheduled caste
Other backward caste
Forward caste
Muslim
Christian
Source: Author’s calculation from the NFHS.
Caste, Culture, and the Status and Well-Being of Widows in India
Table 11.3
Expenditure per capita, body mass index (BMI) and caste, widows and
nonwidows, SARI
Expenditure per capita
Scheduled tribe
Scheduled caste
Other backward caste
Forward caste
363
BMI
Nonwidows
Widows
Nonwidows
Widows
323
(193)
424
(131)
593
(110)
715
(198)
251
(213)
348
(159)
502
(201)
577
(228)
18.9
(.48)
19.3
(.32)
19.5
(.32)
20.6
(.58)
19.1
(.57)
19.4
(.38)
19.3
(.34)
19.4
(.51)
Source: Author’s calculations from SARI data.
gious groups, employment rates are higher among women currently widowed, for both landed and landless households. Over one-half of widowed
scheduled tribe women work, compared to only one-third among the forward castes (though, in percentage terms, the difference between nonwidows and widows is greater for the forward castes). Of course, selection into
widowhood is nonrandom,6 and that selection may differ across these
groups, and we don’t have panel data to examine the prewidow employment rates of current widows. Therefore, these results should only be seen
as descriptive and suggestive.
For a more in-depth investigation of the relationship between caste and
well-being, we apply data from the Survey of Aging in Rural India (SARI),
a survey of 1,477 households, each containing a person aged fifty or older,
conducted by the author in 2002 in five states or Union Territories (Delhi,
Jharkhand, Haryana, Kerala, and Tamil Nadu).7 Several sections of this
survey were modeled to be compatible with other demographic surveys for
India, including the NFHS. The survey collected information on a range of
(current and past) demographic, health, social, and economic variables. In
addition, a village-level survey was conducted in each sample cluster, gathering information on economic and social conditions and infrastructure,
as well as the prevailing customs and norms relating to the elderly.
Table 11.3 shows data on household expenditure8 per capita (with no
adjustments for economies of scale or adult equivalence) for widows and
6. Selection into ownership of land and retention of land after widowhood is also nonrandom.
7. The sample of households for the survey was selected in two stages. In the first stage, village administrative areas, or panchayat, were selected at random from district lists. In the second stage, within each panchayat households were chosen through a random sampling procedure based on registration lists. There was an average of about fifteen households in each
of 103 sampling units selected.
8. Land-owning households were asked the quantity and market value of any consumption
from own production.
364
Robert Jensen
nonwidows in the SARI sample. Across all caste groups, expenditure per
capita is lower for widows than nonwidows, with differences of 20 to 25
percent (though, again, selection into widowhood is nonrandom, and it is
likely that currently widowed women had lower prewidow expenditures
than nonwidows). And across both widows and nonwidows, there is a clear
gradient in expenditure per capita, with scheduled tribes worst off on average, having expenditures per capita less than one-half of that among forward castes.
However, as mentioned above, such data are not entirely indicative of the
consumption or well-being of individuals within households. Instead, we
explore a reduced-form measure of consumption and nutritional status by
examining the body mass index (BMI),9 constructed from physical measurements conducted for all respondents in the survey. The last two columns of table 11.3 provide data on BMI across caste groups. Focusing first
on nonwidows, there is a clear gradient in BMI by caste that mirrors that
of expenditure per capita, with the scheduled tribes or castes having the
lowest BMI and the forward castes having the highest (approximately 10
percent higher than the scheduled tribes). However, comparing widows
and nonwidows within each caste group reveals interesting differences. In
particular, for the scheduled tribes and castes, widows are no worse off in
terms of BMI than nonwidows. However, there are large differences between widows and nonwidows for OBCs and especially the forward castes.
While forward-caste widows still have a slightly higher BMI than widows
in the scheduled tribes, they no longer differ from the scheduled castes or
OBCs. There is no longer any discernible gradient with caste, and in fact
none of the across-caste group differences in BMI are statistically significant. In part, this lack of a gradient likely reflects the fact that there are
stronger restrictions on employment among forward-caste women, as suggested by table 11.2, and may also indicate that widows in forward-caste
households are not given as large a share of total household consumption
as they are in lower-caste households. Thus, wealthier, upper-caste widows
are no better off in terms of nutritional status than poorer but more equally
treated lower-caste widows. Of course, the possibility of differential selection into widowhood across these groups cannot be ruled out.10 Table 11.4
presents comparable data on BMI across caste groups for the NFHS, for
all currently not-pregnant widows and nonwidows aged forty-five to fortynine. Similar patterns emerge in the SARI data, though a distinction arises
between landed and nonlanded households. A similar weakening of the
BMI gradient is seen among widows, and among landless widows the gradient is essentially absent.
9. BMI is weight (in kilograms) divided by squared height (in meters).
10. Of course, in all cases, we are unable to distinguish whether the lower consumption
among widows is by choice or the result of an allocation of resources within the household
over which they have no choice.
Caste, Culture, and the Status and Well-Being of Widows in India
Table 11.4
Body mass index, widows and nonwidows aged 45, NFHS
Landed
Scheduled tribe
Scheduled caste
Other backward caste
Forward caste
Muslim
Christian
365
Landless
Nonwidows
Widows
Nonwidows
Widows
20.2
(.13)
20.4
(.12)
21.2
(.09)
22.9
(.10)
21.8
(.19)
21.9
(.20)
20.2
(.31)
20.7
(.33)
20.7
(.24)
21.7
(.26)
22.6
(.55)
21.8
(.46)
20.1
(.18)
19.8
(.12)
20.5
(.12)
21.3
(.18)
20.5
(.20)
21.4
(.24)
20.5
(.47)
20.1
(.31)
20.3
(.34)
20.3
(.44)
20.4
(.67)
20.9
(.50)
Source: Author’s calculations from NFHS data.
11.4 Institutions, Customs, Norms, and Practices
There are many important institutions, behaviors, and practices surrounding widowhood that affect the well-being of widows, such as whether
the widow is discouraged from remarrying, whether she retains full control
of her husband’s land upon his death, and whether she works. While widows’ rights in all of these cases are protected by law, actual outcomes are
likely to be shaped by societal constraints, traditions, and practices. Further, there are a variety of other social practices, customs, attitudes, and beliefs that affect the status and well-being of widows. For example, following the death of their husband, women are often expected to undergo a
period of seclusion (remaining in a remote room in the home), followed by
confinement to the home (or village) for a period of time, as well as permanently changing their diets (in particular, avoiding “heating” foods
(garam, including meat, eggs, alcohol, onions, and garlic) in favor of “cooling” foods (thanda, including yogurt, rice, milk, and honey; Chen 2000).
Widows are often also unwelcome at social events, ceremonies, and rituals,
and avoided socially because they are considered bad luck, in part because
of their association with death. Sati is the most extreme example of such
practices, although the evidence suggests it is extremely uncommon and in
fact has never been very widespread.
The various norms and practices, and more generally the status of the
elderly and widows in particular, vary widely across India. Many authors
have noted that women fare better in the south relative to the north along
many dimensions of well-being, such as infant mortality, life expectancy,
and education (e.g., Visaria 1967; Bardhan 1974; Miller 1981). Two explanations have typically been suggested for these patterns. The first is that
366
Robert Jensen
wheat is the dominant crop in the north whereas rice is dominant in the
south, and the latter allows women to make a greater economic contribution, which in turn affects women’s status. Consistent with a link between
economic status and the treatment of women, Rosenzweig and Schultz
(1982) find that differences in expected lifetime earnings for men and
women are associated with differentials in male-female survival rates
among children in India. The second explanation offered is the cultural
and historical differences between the populations of the north and south
of India. In particular, the northern populations are the descendents of
Aryans that invaded northern India around 1500 BCE, whereas the dominant population in the south consists of descendents of the Dravidians
who were either originally residing in the south or pushed from the north
when the Aryans arrived. It is generally believed that the Dravidians and
Aryans had widely varying social customs, practices, and organization.11
However, this historical-cultural explanation only pushes the analysis one
step back, and it then becomes important to ask why the Dravidians and
Aryans had differing norms and customs (or why those norms and practices evolved differently once the Aryans arrived).
Our motivating hypothesis focuses around the first explanation, namely
that there are certain economic conditions that differentially affect the
comparative advantage of men and women (and the elderly and the
young). For example, there are some crops, such as rice and tea, that are
less strength intensive than others, such as wheat and other ploughed
crops. Because women and the elderly are able to make a larger economic
contribution in areas growing less strength-intensive crops, we might expect their status to be greater.12 (Though it is also important to take into
consideration that some crops, although strength intensive in their planting and harvesting, potentially provide secondary employment opportunities for women and the elderly in processing those crops, such as the grinding of wheat into flour).13 It should be emphasized that we are not arguing
for a model of “norms” in the sense of a set of behaviors or attitudes based
on shared social beliefs and enforced through social and emotional sanctions and rewards, rather than as part of self-interested behavior; the
greater status of women and the elderly could arise through a model of bar11. For example, it is believed that the Aryans brought caste with them to India, though it
is also possible that the Aryans created caste at the time of their arrival in India, perhaps as a
system to place the Aryans in positions of advantage and prevent intermarriage. Unfortunately, there are no recorded sources to determine whether the Aryans had different attitudes
toward women at the time they invaded or whether those attitudes evolved afterward.
12. We are distinguishing the comparative advantage women or the elderly may have in particular crops from the idea of whether crops are “gendered.” For example, especially in West
Africa, some crops are considered female crops and others male crops. It often has less to do
with the nature of the crop than with whether the crop is grown for own-household consumption (female crops) or as a cash crop (male crops).
13. Of course, it must also be kept in mind that norms could evolve the other way around,
such that norms of respect and care for the elderly arise in places where they are less able to
fend for themselves or live independently.
Caste, Culture, and the Status and Well-Being of Widows in India
367
gaining within the household.14 Empirically, the two are difficult to distinguish. Therefore, we do not argue in favor of one interpretation over the
other, but instead focus on the reduced-form evidence relating the status of
widows to the strength intensity of locally grown crops.
To test this hypothesis, we will explore how various indicators of the status and well-being of widows vary with indicators for whether the dominant crop grown in a village is “female” in the sense that women and the
elderly have a comparative advantage in these crops over other crops.
While the choice is somewhat arbitrary,15 the following crops were considered not conducive to women or the elderly: wheat, maize, and other cereals. Crops considered more amenable to women or the elderly include rice,
tea, tobacco, cotton, and pulse.
Regarding the status of widows, in addition to measures of consumption
and health, the SARI gathered information from respondents regarding
institutions, practices, and customs relating to women, the elderly, and
widows. For example, respondents were asked whether they feel welcome
at ceremonies or “connected” to the society around them. Some questions
regarding status and treatment were asked retrospectively of widows, regarding changes in status and treatment around the time of widowhood.
Our empirical analysis surrounds two primary specifications. We first examine indicators of women’s well-being that vary as a function of widowhood in male versus female crop areas, estimating:
(1)
Wi 0 1Widowi 2 FEMCROPi 3Widowi FEMCROPi
∑ cCASTE i X i i ,
c
where Wi is an indicator of status or well-being and X i is a vector of individual, household and village level covariates. The coefficient of interest
is 3 , the interaction term on widow and female crop—in other words,
whether the difference in Wi for widows and nonwidows differs across male
and female crop areas.16 As indicators of welfare, we use the BMI, selfassessed health status (1 to 5, with higher numbers indicating better
14. In particular, where higher threat point utilities of women or the elderly improve their
allocation within the household. It should be noted that threat points are often considered as
the utility attainable if an individual left the household. While divorce is not common in India, it is not necessary to invoke leaving the household, since it could be that the woman instead threatens not to work or to work at less than her capacity.
15. It is possible to define crops that provide greater female employment opportunities as
those in which we see a larger percent of female labor. However, that strategy introduces an
element of circularity, since working outside of the household is in itself an indicator of
women’s status.
16. In order to allay concerns that crops are chosen in order to accord with the status of
women, in additional specifications we use historical means and variability of rainfall and
temperature as well as soil properties (pH, potassium, nitrogen, phosphorous, salinity, stoniness, and soil type) as instruments for crops grown in the village. However, we cannot rule out
the possibility that groups migrate to places where the nature of agriculture and the opportunities for women are more consistent with their preexisting views on women’s status.
368
Robert Jensen
health), whether they feel they are respected by others in the community,
and how connected to society they feel (the last two of which also range
from 1 to 5, with higher numbers indicating more respect and more connection). We also include a set of caste dummy variables (scheduled tribe,
scheduled caste, and OBC), and indicators for whether the widow lives
alone (about 11 percent) or with her children (about 56 percent).17 As additional covariates, we also include age, education, income, household size,
and whether the respondent is the head of the household. As stated, this is
a difference-in-differences estimator, comparing the difference in the outcome of interest between widows and nonwidows in female crop areas relative to the difference in male crop areas. Unfortunately, the data are crosssectional, and it is not possible to track individuals over time or examine
selection into widowhood (which in itself could differ across male and female crop areas). Our goal will simply be to focus on whether the differences between widows and nonwidows vary with the crops grown locally.
Further, we are only examining differences upon entering into widowhood
in these areas: it is likely that women may be treated worse in male crop areas even when they are not widows, in which case we might not expect a
large differential change at widowhood.
To examine the practices and changes in well-being, treatment, and status surrounding widowhood in male versus female crop areas, we estimate
(2)
Wi 0 1FEMCROPi ∑ cCASTE i Xi i ,
c
which has the advantage of explicitly examining changes in treatment and
status in the two types of areas, rather than comparing widows and nonwidows. As indicators of changes in welfare following widowhood we examine the following: whether, at the time of their husband’s death, there
was change in treatment by the husband’s family (1: better, 2: about the
same, 3: worse); whether they lost control of their husband’s land (0,1 conditional on having owned land); whether they noticed a change in their status among other people in the village (1: better, 2: about the same, 3:
worse); whether they feel less welcome at ceremonies and events such as
weddings or births (0,1); how long they remained inside the household after the death of their husband (in months); and whether they changed their
diet (0,1) toward thanda foods. While these are imperfect measures and do
not capture the full complexity of well-being, they are useful proxies for
treatment and status along key dimensions following widowhood.
The results of regression (1) are presented in table 11.5. Across all four
measures, BMI, self-assessed health, respect from others in the village, and
sense of connectedness, widows are worse off than nonwidows. Body mass
17. Of the remainders, about 30 percent live with their deceased husband’s family, 3 percent
with other people.
Caste, Culture, and the Status and Well-Being of Widows in India
Table 11.5
Change in status and well-being indicators, widows vs.
nonwidows, NFHS
Widow
Female crop
Female crop widow
Age
Education
Scheduled tribe
Scheduled caste
Other backward caste
North
369
BMI
(1)
Self-assessed
health
(2)
Respect from
others in village
(3)
Connectedness
(4)
–1.10
(.49)
–.29
(.16)
.69
(.21)
–.008
(.005)
.14
(.01)
–.68
(.16)
–.79
(.15)
–.34
(.15)
–.27
(.12)
–.29
(.15)
–.07
(.10)
.24
(.13)
–.005
(.003)
.13
(.01)
–.31
(.10)
–.39
(.09)
–.27
(.09)
–.10
(.07)
–.65
(.27)
–.27
(.08)
.30
(.11)
–.003
(.003)
.021
(.007)
.34
(.08)
.35
(.07)
.29
(.07)
.08
(.06)
–.56
(.31)
–.12
(.06)
.48
(.19)
–.01
(.01)
.16
(.05)
.57
(.06)
.56
(.05)
.33
(.05)
.08
(.04)
Notes: All results from the SARI data set. Sample restricted to women. The number of observations is 1,249. The dependent variable in column (1) is the body mass index (BMI). Selfassessed health status in column (2) is an indicator ranging from one to five (1 very bad, 2
bad, 3 about average, 4 good, 5 very good). Respect from others in the village and
connectedness also range from 1 to 5, with higher numbers indicating more respect and more
connection to the village.
index is also statistically significantly higher in female crop areas, though
this may in part not reflect an intrahousehold allocation issue as much as
the fact that areas with male crops tend to be slightly better off on average.
But in all four cases, the interaction term on female crop and widow suggests that the difference between widows and nonwidows is much larger in
areas that do not have female crops. In fact, for self-assessed health and the
sense of connectedness with the community, while widows are worse off in
male crop areas, we cannot reject the hypothesis that widows are no worse
off than nonwidows in female crop areas. Women in the lower castes generally have lower BMI and self-assessed health, as would be expected given
the socioeconomic differentials by caste. However, interestingly, lowercaste groups feel more respect and social connectedness, which seems inconsistent with the view that lower castes explicitly have lower status under the caste system. However, it is possible that lower castes feel more
respect or connection to others in the village of the same caste, since residence is often segregated by caste; unfortunately, the question did not
make this distinction or specify the group meant by “the village.”
370
Robert Jensen
Table 11.6
Female crop
Scheduled tribe
Scheduled caste
Other backward
caste
Education
Comparative change in status and well-being indicators, widows
Treatment
husb. family
(1)
Diet
(2)
Lost land
(had before)
(3)
Status
in village
(4)
Less
welcome
(5)
How long
stay inside
(6)
–.31
(.15)
–.25
(.06)
–.19
(.05)
–.04
(.05)
–.01
(.006)
–.18
(.06)
–.06
(.05)
–.10
(.04)
–.13
(.04)
.002
(.004)
–.21
(.08)
–.12
(.07)
–.03
(.06)
–.09
(.06)
–.001
(.001)
–.16
(.06)
–.10
(.05)
–.08
(.04)
.005
(.04)
–.005
(.004)
–.13
(.07)
–.039
(.05)
–.034
(.04)
.018
(.04)
–.002
(.004)
–.94
(.39)
–3.2
(.27)
–3.1
(.21)
–1.7
(.28)
–.13
(.02)
Notes: All results from the SARI data set. Sample restricted to ever-widowed women. The number of observations is 978 for all columns except column (5), where it is 453 because the sample is restricted to those
who owned land before widowhood and who have not remarried (since nearly all of those who remarry
report losing their original land). In column (1), the dependent variable is whether the respondent feels
her husband’s family treated her 1) better, 2) about the same, or 3) worse following the death of her husband. In column (2), it is an indicator for whether the wife changed her diet following the death of her husband. In column (4), it is whether other people in the village treated her 1) better, 2) about the same, or 3)
worse following the death of her husband. In column (3), the dependent variable is whether the widow lost
access to her husband’s land after his death. In column (5), it is an indicator for whether she felt less welcome at ceremonies such as weddings or births, and in column (6), the dependent variable is how long (in
months) the woman stayed inside her husband’s household following his death ceremony.
Table 11.6 further confirms that there are fewer adverse consequences of
widowhood in female crop areas relative to male crop areas, using widows’
self-reports of changes in indicators of welfare. Widows are less likely to report having experienced a decline in their relationship with their husband’s
family following the death of their husband, and among those who have
not remarried, they are less likely to have lost their land (typically, to the
family of the husband) in cases where they had owned land. In terms of social connectedness, they also are less likely to report feeling less welcome at
public events or having experienced an overall decline in their status in the
village since becoming widowed. Finally, in terms of individual behavior,
widows are significantly less likely to have changed their diet in female crop
areas (which could also account for some of the smaller decline in BMI in
female relative to male crop areas) and spent on average about a month less
confined to their deceased husband’s home following his death.
The results on caste are also consistent with the view that widows are
treated better and have higher status among the lower castes relative to the
higher castes. In particular, relative to widows from forward castes, widows
from scheduled tribe and scheduled caste groups are significantly less likely
to have experienced a decline in the relationship with their husband’s family or their status in the village upon widowhood. They are also less likely
to have changed their diet and remained in their husband’s home for significantly less time.
Caste, Culture, and the Status and Well-Being of Widows in India
371
11.5 Conclusion
There are two primary lessons from this paper. The first is that issues of
intrahousehold allocation are essential for assessing individual well-being,
especially for the elderly. While this has been widely appreciated in
economics for some time, very little is done about it in practice, and most
studies focus only on household per capita measures in assessing living
standards. This observation also has implications for studies on the relationship between SES and health. Further, the results also show that the relationship between SES and health or nutrition in India is more complex
than simply the purchasing power potentially implied by income or expenditure; in particular, other factors, such as the treatment of individuals
within the household, mediate this relationship. In particular, we find that
widows are much better off in forward-caste households when measured in
terms of per capita expenditure, but when BMI is used as a crude proxy for
consumption of the elderly, forward-caste persons are no better off than
lower-caste households, suggesting the share of household resources is not
well proxied by expenditure per person.
The second lesson of this paper is that the status, treatment, and wellbeing of widows have a foundation in potential economic value, either
through bargaining power within households or through a cultural underpinning to the evolution of cultural norms. Of course, this is not to claim
that economic factors are the only, or even the largest, determinant of “culture” or to argue that factors such as history and lineage are not important.
However, the evidence indicates that economic factors do appear to play at
least some role in influencing the well-being and status of widows. The implication is that programs such as microenterprise ventures that expand
economic opportunities for women or the elderly, attempts to minimize age
or gender discrimination in private-sector employment, or gender- or agesensitive hiring schemes for public projects, especially in places where the
state employs a significant number of people, may improve women’s and
widows’ status.
Appendix
A Brief Note on Caste in India 18
Throughout India, caste is a social structure that is pervasive, highly visible, and an influence on most aspects of life, including employment and
social opportunities, access to resources, customs, diet, occupation, and
relations with others. Caste is transmitted through birth, though it is com18. This section draws broadly from Dumont (1980).
372
Robert Jensen
plex, dynamic, and changing: through marriage or changing some activities or practices, some castes move up or down. In general, there is a great
deal of caste segregation regarding marriage and residential patterns.
The system referred to as caste today has its roots in the earlier varna
(color) system. In the early Vedic period, society was said to be organized
into four broad social groups, called varnas. The varnas represented a social hierarchy based on notions of purity and pollution. Some kinds of
work are pure and some are impure or polluted, and the varna system was
organized around the principle of delegating the various activities to particular groups, in particular to allow religious clerics to remain ritualistically pure. At the top of the hierarchy are the Brahmins (priests and teachers); then come the Kshatriyas (rulers and warriors), the Vaishras (traders
and farmers), and the Shudras (servants or menial workers; more commonly referred to today as other backward castes [OBCs]). There is a fifth
group outside this categorization, called by a variety of names such as the
Outcastes, Dalits, Harijans, Untouchables, or the Scheduled Castes. For
India, 15 percent of the population comes from the upper or forward
castes, 15 percent are scheduled castes, and 8 percent scheduled tribes,
with the remaining bulk of the population the Shudras or OBCs. Over time,
the varna system evolved into or combined with the jati system, which is
what is generally meant today by caste. Jatis are smaller categories that
vary widely from place to place in India and can roughly be categorized to
correspond to the varnas (or the scheduled caste category). The system has
evolved over time, and there currently are over 3,000 castes and 20,000 subcastes, with a fairly well-defined social ranking.
References
Agarwala, S. N. 1970. A demographic study of six urbanising villages. Bombay: Asia
Publishing House.
Bardhan, Pranab. 1974. On life and death questions. Economic and Political Weekly
9:1293–1304.
Bhat, M., and R. Kanbargi. 1984. Estimating the incidence of widow and widower
re-marriage using census data. Population Studies 3 (1): 89–103.
Chen, Martha Alter. 2000. Perpetual mourning: Widowhood in rural India. New
Delhi: Oxford University Press.
Chen, Martha Alter, and Jean Drèze. 1995. Widowhood and well-being in rural
North India. In Women’s health in India: Risk and vulnerability, ed. Monica Das
Gupta, Lincoln Chen, and T. N. Krishnan, 245–88. New Delhi: Oxford University Press.
Deaton, Angus, and Christina Paxson. 1998. Measuring poverty among the elderly.
In Inquiries in the economics of aging, ed. David Wise, 169–200. Chicago: University of Chicago Press.
Drèze, Jean. 1990. Widows in rural India. Department of Economics Research
Program paper no. 26. London: London School of Economics.
Caste, Culture, and the Status and Well-Being of Widows in India
373
Drèze, Jean, and P. V. Srinivasan. 1997. Widowhood and poverty in rural India:
Some inferences from household survey data. Journal of Development Economics 54:217–34.
Dumont, Louis. 1980. Homo heirarchicus: The caste system and its implications.
Chicago: University of Chicago Press.
Irudaya Rajan, S., U. S. Mishra, and P. Sankara Sarma. 1999. India’s elderly: Burden or challenge? New Delhi: Sage Publications.
Miller, Barbara D. 1981. The endangered sex: Neglect of female children in rural
North India. Ithaca, New York: Cornell University Press.
Rosenzweig, Mark, and T. Paul Schultz. 1982. Market opportunities, genetic endowments and intrafamily resource distribution: Child survival in rural India.
American Economic Review 72 (4): 803–15.
Sen, Amartya. 1999. Development as freedom. New York: Anchor Books.
United Nations. 1961. The Mysore population study. New York: United Nations.
Visaria, Pravin M. 1967. The sex ratio of the population of India. Census of India
1961, Vol. 1, Monograph #10. New Delhi.
Comment
Esther Duflo
The plight of widows in India is the focus of a growing literature. Widows
are eminently vulnerable for several reasons: they are women; they are often relatively old (although women who married older men and did not remarry after their spouse’s death can be young widows); and norms and usages restrict their mobility, the exercise of their property rights, and even
the food they can eat and the dress they can wear.
This paper uses existing and original data from India to cast new light
on two important issues: whether the treatment of widows is affected by
their castes, and whether it is affected by their potential “usefulness.” The
main result of the first part is that, while widows from “forward” castes live
in richer households than those in “backward” castes (which reflects the
general correlation between caste and economic status), their own health
status, reflected in their BMI, is no higher in forward than in backward
castes. This is not true for nonwidows, who have a higher BMI in households from richer castes. The author interprets this evidence as suggesting
that widows in high-caste households actually get a smaller share of a bigger cake, reflecting the more conservative habits of so-called forward castes.
This is not the only possible interpretation, however: the sample does not
seem to be restricted to older women, since, at least in the NFHS, most
measurements are only obtained for women aged fifteen to forty-nine. If it
is not, it might be that the differential caste gradient for widows and nonwidows just reflects a differential caste gradient for older women than for
younger women, perhaps indicating a differential impact of food intake on
Esther Duflo is professor of Economics at the Massachusetts Institute of Technology and
a research associate of the National Bureau of Economic Research.
374
Robert Jensen
BMI in older women, or that older women develop different priorities as
they grow older, so that when they are better treated, they feed their grandchildren more. For example, Duflo (2003) shows that when the bargaining
power of grandmothers increased in South Africa (as they received a substantial old age pension), their granddaughters were better fed. More information would be useful in sorting out some of Jensen’s evidence: What
is the caste gradient in self-reported health measures (used later in the paper)? What is the income gradient for widows and nonwidows? Does the
differential caste gradient persist even after controlling for age? Does the
income gradient differ by caste group? Is there a caste gradient in reported
norms and usages?
The second part of the paper seeks to explain differences in the treatment of widows across regions. It hypothesizes that widows would be better treated in regions where women are more productive. The main result
is that widows are better treated, relative to nonwidows, in places that grow
rice (or other crops intensive in “women’s labor”) than in places that grow
wheat (or other crops intensive in “men’s labor”). This shows up in two
types of variables: direct measure of health (BMI and self-reported health
status) and report of constraints (on food, dress, mobility) associated with
becoming a widow.
There is a bit of a theoretical lapse between the initial argument and the
actual results: the argument involves the comparative advantage of women
in general, not of widows in particular. Yet the regressions show that widows are doing better in regions where women have a comparative advantage relative to nonwidows. This is somewhat surprising, in view of the initial argument, since one would think that widows are less productive than
other women and thus lose part of their comparative advantage vis-à-vis
men. It is also very difficult to ignore the fact that the comparison essentially boils down to a comparison of different types of regions: as the author himself points out, there is a little bit of circularity in defining which
crops are naturally more suited to women than to men, and to take the crop
mix (or even the geographical suitability) as exogenous in a regression that
seeks to explain longstanding customs. It might well be that communities
that treat women more equally have evolved technology for production of
crops which gives an important role to women. Perhaps if Dravidians had
lived in the north instead of the south, wheat growing would actually be female intensive and rice would not. Instrumenting with weather or suitability would address this issue.
In this context, the positive difference-in-difference becomes more troubling than reassuring: since there are no theoretical reasons to expect one
(in fact, a technological argument would lead one to expect the opposite),
perhaps it comes precisely from these omitted variables. This is made even
more troubling by the fact that, in general, women actually do not seem to
be better off in rice-producing regions.
Caste, Culture, and the Status and Well-Being of Widows in India
375
That being said, the main insight of the paper, that the treatment of widows, even if ostensibly governed by timeless norms, may actually respond
in part to economic motives, is very important. The evidence provided
here, while far from definitive, is certainly tantalizing. Chen and Drèze
(1995) provide related evidence when they report that the presence of a
widow’s pension protects the widow from poor treatment. Abhijit Banerjee, Angus Deaton, and I hope to test this idea directly: widows in India are
eligible for a pension, subject to certain conditions. The Panchayat is in
charge of identifying potential beneficiaries. However, very few potentially
eligible people actually receive the pension. This is probably due in part to
the difficulties of navigating the system, establishing one’s rights, and following through until the pension is actually delivered. We are planning to
evaluate the work of Seva Mandir, a local nongovernmental organization,
in one rural district in Rajasthan that will provide assistance to eligible
widows to receive their pension. By comparing the treatment of widows
who have received this assistance to those who have not, we might be able
to provide a further test of the very interesting hypothesis brought forward
in this paper.
References
Chen, Martha, and Jean Drèze. 1995. Widowhood and well-being in rural North
India. In Women’s health in India: Risk and vulnerability, ed. Monica Das Gupta,
Lincoln Chen, and T. N. Krishnan, 245–88. New Delhi: Oxford University Press.
Duflo, Esther. 2003. Grandmothers and granddaughters: Old age pension and
intra-household allocation in South Africa. World Bank Economic Review 17 (1):
1–25.